Fix compiler issues again

.github/workflows/ci.yaml

@@ -47,7 +47,7 @@ jobs:
 
       matrix:
 
-        # Need dummy entries for the matrix to pick up the variables.
+        # Add basic entries for the matrix to run the latest versions.
         # We later set all other combinations that we want explicitly.
         os:
           - ubuntu-latest
@@ -68,7 +68,23 @@ jobs:
 
         # C++ standard
         cppstd:
-           - auto
+          - auto
+
+        # -------------------------------------------------------
+        #     exclude
+        # -------------------------------------------------------
+
+        exclude:
+
+          # Currently (as of 2024-12), the latest llvm used here is clang 18,
+          # which has several bugs concerning deprecated functions, see
+          # https://github.com/llvm/llvm-project/issues/76515 and
+          # https://gcc.gnu.org/pipermail/libstdc++/2024-March/058502.html
+          # which only seem to occur when the environment also has a gcc std lib
+          # present. That is mostly a setup issue here in the CI, so we do not
+          # solve this for now, and just deactivate that test case completely.
+          - os: ubuntu-latest
+            compiler: llvm
 
         # -------------------------------------------------------
         #     include
@@ -87,7 +103,8 @@ jobs:
         # the very same version of llvm-9, and got tons of segfaults for all kind of functions.
 
         # We also currently exclude clang from macos, as it fails with different types of
-        # linker errors. As this is merely a problem in the setup, we defer solving this for now.
+        # linker errors. As this is merely a problem in the setup in the CI environment,
+        # we defer solving this for now.
 
         include:
 
@@ -449,9 +466,10 @@ jobs:
           - os: ubuntu-latest
             compiler: gcc
             htslib: OFF
-          - os: ubuntu-latest
-            compiler: llvm
-            htslib: OFF
+          # Deactivate due to bug in llvm 18, see above
+          # - os: ubuntu-latest
+          #   compiler: llvm
+          #   htslib: OFF
           - os: macos-latest
             compiler: gcc
             htslib: OFF

lib/genesis/utils/threading/thread_pool.hpp

@@ -496,9 +496,11 @@ class ThreadPool
         // All this wrapping should be completely transparent to the compiler, and removed.
         // The task captures the package including the promise that is needed for the future.
         WrappedTask wrapped_task;
+        std::function<result_type()> task_function = std::bind(
+            std::forward<F>(f), std::forward<Args>(args)...
+        );
         wrapped_task.function = make_wrapped_task_with_promise_(
-            task_promise,
-            make_task_function_( std::forward<F>(f), std::forward<Args>(args)... )
+            task_promise, std::move( task_function )
         );
 
         // We first incrementi the unfinished counter, and only decrementing it once the task has
@@ -531,7 +533,7 @@ class ThreadPool
         // All this wrapping should be completely transparent to the compiler, and removed.
         // The task captures the package including the promise that is needed for the future.
         WrappedTask wrapped_task;
-        auto task_function = make_task_function_( std::forward<F>(f), std::forward<Args>(args)... );
+        auto task_function = std::bind( std::forward<F>(f), std::forward<Args>(args)... );
         wrapped_task.function = [task_function, this]()
         {
             // Run the actual work task here. Once done, we can signal this to the unfinished list.
@@ -698,35 +700,6 @@ class ThreadPool
         }
     }
 
-    template<typename F, typename... Args>
-    inline auto make_task_function_( F&& f, Args&&... args )
-    -> std::function<typename genesis_invoke_result<F, Args...>::type ()>
-    {
-        // Unfortunately, Clang 18 when compiled under the C++20 standard somehow uses
-        // std::result_of within std::bind, despite that being deprecated in the standard,
-        // and hence leads to a warning, and as we set warnings as errors, fails to compile.
-        // See https://gcc.gnu.org/pipermail/libstdc++/2024-March/058502.html for details.
-        // This is the reason why we internally use genesis_invoke_result instead, to switch
-        // between the two. But doesn't work of course for the STL...
-        // So we need a workaround for this. Silencing via #pragma diagnostic does not seem
-        // to work, so instead, we completely get rid of the std::bind and use perfect captures.
-        // This is the whole reason for this function. Super ugly, but it is what it is.
-        #if GENESIS_CPP_STD >= GENESIS_CPP_STD_20
-
-            // Use a modern way to bind the args to the function.
-            return [f = std::forward<F>(f), ...args = std::forward<Args>(args)]() mutable
-            {
-                return f(args...);
-            };
-
-        #else
-
-            // Make the function via binding.
-            return std::bind( std::forward<F>(f), std::forward<Args>(args)... );
-
-        #endif
-    }
-
     template<typename T>
     inline std::function<void()> make_wrapped_task_with_promise_(
         std::shared_ptr<std::promise<T>>& task_promise,

test/src/utils/threading/concurrent_vector_guard.cpp

@@ -117,7 +117,7 @@ TEST( Threading, VectorEntries )
     // them from their vec within data, until empty, and then moves to the next one.
     auto thread_pool = std::make_shared<ThreadPool>( num_threads );
     auto vector_guard = ConcurrentVectorGuard( num_vecs );
-    auto num_elem = std::atomic<size_t>{0};
+    std::atomic<size_t> num_elem{0};
     for( size_t t = 0; t < num_threads; ++t ) {
         thread_pool->enqueue_detached([&, t]()
         {